2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * See Documentation/technical/api-directory-listing.txt
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
13 #include "wildmatch.h"
16 struct path_simplify
{
22 * Tells read_directory_recursive how a file or directory should be treated.
23 * Values are ordered by significance, e.g. if a directory contains both
24 * excluded and untracked files, it is listed as untracked because
25 * path_untracked > path_excluded.
34 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
35 const char *path
, int len
,
36 int check_only
, const struct path_simplify
*simplify
);
37 static int get_dtype(struct dirent
*de
, const char *path
, int len
);
39 /* helper string functions with support for the ignore_case flag */
40 int strcmp_icase(const char *a
, const char *b
)
42 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
45 int strncmp_icase(const char *a
, const char *b
, size_t count
)
47 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
50 int fnmatch_icase(const char *pattern
, const char *string
, int flags
)
52 return fnmatch(pattern
, string
, flags
| (ignore_case
? FNM_CASEFOLD
: 0));
55 inline int git_fnmatch(const char *pattern
, const char *string
,
56 int flags
, int prefix
)
59 if (flags
& GFNM_PATHNAME
)
60 fnm_flags
|= FNM_PATHNAME
;
62 if (strncmp(pattern
, string
, prefix
))
67 if (flags
& GFNM_ONESTAR
) {
68 int pattern_len
= strlen(++pattern
);
69 int string_len
= strlen(string
);
70 return string_len
< pattern_len
||
72 string
+ string_len
- pattern_len
);
74 return fnmatch(pattern
, string
, fnm_flags
);
77 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
78 const char *string
, int stringlen
,
82 struct strbuf pat_buf
= STRBUF_INIT
;
83 struct strbuf str_buf
= STRBUF_INIT
;
84 const char *use_pat
= pattern
;
85 const char *use_str
= string
;
87 if (pattern
[patternlen
]) {
88 strbuf_add(&pat_buf
, pattern
, patternlen
);
89 use_pat
= pat_buf
.buf
;
91 if (string
[stringlen
]) {
92 strbuf_add(&str_buf
, string
, stringlen
);
93 use_str
= str_buf
.buf
;
98 match_status
= wildmatch(use_pat
, use_str
, flags
, NULL
);
100 strbuf_release(&pat_buf
);
101 strbuf_release(&str_buf
);
106 static size_t common_prefix_len(const char **pathspec
)
108 const char *n
, *first
;
110 int literal
= limit_pathspec_to_literal();
116 while ((n
= *pathspec
++)) {
118 for (i
= 0; first
== n
|| i
< max
; i
++) {
120 if (!c
|| c
!= first
[i
] || (!literal
&& is_glob_special(c
)))
125 if (first
== n
|| len
< max
) {
135 * Returns a copy of the longest leading path common among all
138 char *common_prefix(const char **pathspec
)
140 unsigned long len
= common_prefix_len(pathspec
);
142 return len
? xmemdupz(*pathspec
, len
) : NULL
;
145 int fill_directory(struct dir_struct
*dir
, const char **pathspec
)
150 * Calculate common prefix for the pathspec, and
151 * use that to optimize the directory walk
153 len
= common_prefix_len(pathspec
);
155 /* Read the directory and prune it */
156 read_directory(dir
, pathspec
? *pathspec
: "", len
, pathspec
);
160 int within_depth(const char *name
, int namelen
,
161 int depth
, int max_depth
)
163 const char *cp
= name
, *cpe
= name
+ namelen
;
169 if (depth
> max_depth
)
176 * Does 'match' match the given name?
177 * A match is found if
179 * (1) the 'match' string is leading directory of 'name', or
180 * (2) the 'match' string is a wildcard and matches 'name', or
181 * (3) the 'match' string is exactly the same as 'name'.
183 * and the return value tells which case it was.
185 * It returns 0 when there is no match.
187 static int match_one(const char *match
, const char *name
, int namelen
)
190 int literal
= limit_pathspec_to_literal();
192 /* If the match was just the prefix, we matched */
194 return MATCHED_RECURSIVELY
;
198 unsigned char c1
= tolower(*match
);
199 unsigned char c2
= tolower(*name
);
200 if (c1
== '\0' || (!literal
&& is_glob_special(c1
)))
210 unsigned char c1
= *match
;
211 unsigned char c2
= *name
;
212 if (c1
== '\0' || (!literal
&& is_glob_special(c1
)))
223 * If we don't match the matchstring exactly,
224 * we need to match by fnmatch
226 matchlen
= strlen(match
);
227 if (strncmp_icase(match
, name
, matchlen
)) {
230 return !fnmatch_icase(match
, name
, 0) ? MATCHED_FNMATCH
: 0;
233 if (namelen
== matchlen
)
234 return MATCHED_EXACTLY
;
235 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
236 return MATCHED_RECURSIVELY
;
241 * Given a name and a list of pathspecs, returns the nature of the
242 * closest (i.e. most specific) match of the name to any of the
245 * The caller typically calls this multiple times with the same
246 * pathspec and seen[] array but with different name/namelen
247 * (e.g. entries from the index) and is interested in seeing if and
248 * how each pathspec matches all the names it calls this function
249 * with. A mark is left in the seen[] array for each pathspec element
250 * indicating the closest type of match that element achieved, so if
251 * seen[n] remains zero after multiple invocations, that means the nth
252 * pathspec did not match any names, which could indicate that the
253 * user mistyped the nth pathspec.
255 int match_pathspec(const char **pathspec
, const char *name
, int namelen
,
256 int prefix
, char *seen
)
266 for (i
= 0; pathspec
[i
] != NULL
; i
++) {
268 const char *match
= pathspec
[i
] + prefix
;
269 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
271 how
= match_one(match
, name
, namelen
);
275 if (seen
&& seen
[i
] < how
)
283 * Does 'match' match the given name?
284 * A match is found if
286 * (1) the 'match' string is leading directory of 'name', or
287 * (2) the 'match' string is a wildcard and matches 'name', or
288 * (3) the 'match' string is exactly the same as 'name'.
290 * and the return value tells which case it was.
292 * It returns 0 when there is no match.
294 static int match_pathspec_item(const struct pathspec_item
*item
, int prefix
,
295 const char *name
, int namelen
)
297 /* name/namelen has prefix cut off by caller */
298 const char *match
= item
->match
+ prefix
;
299 int matchlen
= item
->len
- prefix
;
301 /* If the match was just the prefix, we matched */
303 return MATCHED_RECURSIVELY
;
305 if (matchlen
<= namelen
&& !strncmp(match
, name
, matchlen
)) {
306 if (matchlen
== namelen
)
307 return MATCHED_EXACTLY
;
309 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
310 return MATCHED_RECURSIVELY
;
313 if (item
->nowildcard_len
< item
->len
&&
314 !git_fnmatch(match
, name
,
315 item
->flags
& PATHSPEC_ONESTAR
? GFNM_ONESTAR
: 0,
316 item
->nowildcard_len
- prefix
))
317 return MATCHED_FNMATCH
;
323 * Given a name and a list of pathspecs, returns the nature of the
324 * closest (i.e. most specific) match of the name to any of the
327 * The caller typically calls this multiple times with the same
328 * pathspec and seen[] array but with different name/namelen
329 * (e.g. entries from the index) and is interested in seeing if and
330 * how each pathspec matches all the names it calls this function
331 * with. A mark is left in the seen[] array for each pathspec element
332 * indicating the closest type of match that element achieved, so if
333 * seen[n] remains zero after multiple invocations, that means the nth
334 * pathspec did not match any names, which could indicate that the
335 * user mistyped the nth pathspec.
337 int match_pathspec_depth(const struct pathspec
*ps
,
338 const char *name
, int namelen
,
339 int prefix
, char *seen
)
344 if (!ps
->recursive
||
345 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
347 return MATCHED_RECURSIVELY
;
349 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
350 return MATCHED_EXACTLY
;
358 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
360 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
362 how
= match_pathspec_item(ps
->items
+i
, prefix
, name
, namelen
);
364 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
365 ps
->max_depth
!= -1 &&
366 how
&& how
!= MATCHED_FNMATCH
) {
367 int len
= ps
->items
[i
].len
;
368 if (name
[len
] == '/')
370 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
371 how
= MATCHED_EXACTLY
;
378 if (seen
&& seen
[i
] < how
)
386 * Return the length of the "simple" part of a path match limiter.
388 int simple_length(const char *match
)
393 unsigned char c
= *match
++;
395 if (c
== '\0' || is_glob_special(c
))
400 int no_wildcard(const char *string
)
402 return string
[simple_length(string
)] == '\0';
405 void parse_exclude_pattern(const char **pattern
,
410 const char *p
= *pattern
;
415 *flags
|= EXC_FLAG_NEGATIVE
;
419 if (len
&& p
[len
- 1] == '/') {
421 *flags
|= EXC_FLAG_MUSTBEDIR
;
423 for (i
= 0; i
< len
; i
++) {
428 *flags
|= EXC_FLAG_NODIR
;
429 *nowildcardlen
= simple_length(p
);
431 * we should have excluded the trailing slash from 'p' too,
432 * but that's one more allocation. Instead just make sure
433 * nowildcardlen does not exceed real patternlen
435 if (*nowildcardlen
> len
)
436 *nowildcardlen
= len
;
437 if (*p
== '*' && no_wildcard(p
+ 1))
438 *flags
|= EXC_FLAG_ENDSWITH
;
443 void add_exclude(const char *string
, const char *base
,
444 int baselen
, struct exclude_list
*el
, int srcpos
)
451 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
452 if (flags
& EXC_FLAG_MUSTBEDIR
) {
454 x
= xmalloc(sizeof(*x
) + patternlen
+ 1);
456 memcpy(s
, string
, patternlen
);
457 s
[patternlen
] = '\0';
460 x
= xmalloc(sizeof(*x
));
463 x
->patternlen
= patternlen
;
464 x
->nowildcardlen
= nowildcardlen
;
466 x
->baselen
= baselen
;
469 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
470 el
->excludes
[el
->nr
++] = x
;
474 static void *read_skip_worktree_file_from_index(const char *path
, size_t *size
)
478 enum object_type type
;
480 struct index_state
*istate
= &the_index
;
483 pos
= index_name_pos(istate
, path
, len
);
486 if (!ce_skip_worktree(istate
->cache
[pos
]))
488 data
= read_sha1_file(istate
->cache
[pos
]->sha1
, &type
, &sz
);
489 if (!data
|| type
!= OBJ_BLOB
) {
498 * Frees memory within el which was allocated for exclude patterns and
499 * the file buffer. Does not free el itself.
501 void clear_exclude_list(struct exclude_list
*el
)
505 for (i
= 0; i
< el
->nr
; i
++)
506 free(el
->excludes
[i
]);
515 int add_excludes_from_file_to_list(const char *fname
,
518 struct exclude_list
*el
,
522 int fd
, i
, lineno
= 1;
526 fd
= open(fname
, O_RDONLY
);
527 if (fd
< 0 || fstat(fd
, &st
) < 0) {
529 warn_on_inaccessible(fname
);
533 (buf
= read_skip_worktree_file_from_index(fname
, &size
)) == NULL
)
539 if (buf
[size
-1] != '\n') {
540 buf
= xrealloc(buf
, size
+1);
545 size
= xsize_t(st
.st_size
);
550 buf
= xmalloc(size
+1);
551 if (read_in_full(fd
, buf
, size
) != size
) {
562 for (i
= 0; i
< size
; i
++) {
563 if (buf
[i
] == '\n') {
564 if (entry
!= buf
+ i
&& entry
[0] != '#') {
565 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
566 add_exclude(entry
, base
, baselen
, el
, lineno
);
575 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
576 int group_type
, const char *src
)
578 struct exclude_list
*el
;
579 struct exclude_list_group
*group
;
581 group
= &dir
->exclude_list_group
[group_type
];
582 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
583 el
= &group
->el
[group
->nr
++];
584 memset(el
, 0, sizeof(*el
));
590 * Used to set up core.excludesfile and .git/info/exclude lists.
592 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
594 struct exclude_list
*el
;
595 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
596 if (add_excludes_from_file_to_list(fname
, "", 0, el
, 0) < 0)
597 die("cannot use %s as an exclude file", fname
);
600 int match_basename(const char *basename
, int basenamelen
,
601 const char *pattern
, int prefix
, int patternlen
,
604 if (prefix
== patternlen
) {
605 if (patternlen
== basenamelen
&&
606 !strncmp_icase(pattern
, basename
, basenamelen
))
608 } else if (flags
& EXC_FLAG_ENDSWITH
) {
609 /* "*literal" matching against "fooliteral" */
610 if (patternlen
- 1 <= basenamelen
&&
611 !strncmp_icase(pattern
+ 1,
612 basename
+ basenamelen
- (patternlen
- 1),
616 if (fnmatch_icase_mem(pattern
, patternlen
,
617 basename
, basenamelen
,
624 int match_pathname(const char *pathname
, int pathlen
,
625 const char *base
, int baselen
,
626 const char *pattern
, int prefix
, int patternlen
,
633 * match with FNM_PATHNAME; the pattern has base implicitly
636 if (*pattern
== '/') {
643 * baselen does not count the trailing slash. base[] may or
644 * may not end with a trailing slash though.
646 if (pathlen
< baselen
+ 1 ||
647 (baselen
&& pathname
[baselen
] != '/') ||
648 strncmp_icase(pathname
, base
, baselen
))
651 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
652 name
= pathname
+ pathlen
- namelen
;
656 * if the non-wildcard part is longer than the
657 * remaining pathname, surely it cannot match.
659 if (prefix
> namelen
)
662 if (strncmp_icase(pattern
, name
, prefix
))
665 patternlen
-= prefix
;
670 * If the whole pattern did not have a wildcard,
671 * then our prefix match is all we need; we
672 * do not need to call fnmatch at all.
674 if (!patternlen
&& !namelen
)
678 return fnmatch_icase_mem(pattern
, patternlen
,
684 * Scan the given exclude list in reverse to see whether pathname
685 * should be ignored. The first match (i.e. the last on the list), if
686 * any, determines the fate. Returns the exclude_list element which
687 * matched, or NULL for undecided.
689 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
691 const char *basename
,
693 struct exclude_list
*el
)
698 return NULL
; /* undefined */
700 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
701 struct exclude
*x
= el
->excludes
[i
];
702 const char *exclude
= x
->pattern
;
703 int prefix
= x
->nowildcardlen
;
705 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
706 if (*dtype
== DT_UNKNOWN
)
707 *dtype
= get_dtype(NULL
, pathname
, pathlen
);
708 if (*dtype
!= DT_DIR
)
712 if (x
->flags
& EXC_FLAG_NODIR
) {
713 if (match_basename(basename
,
714 pathlen
- (basename
- pathname
),
715 exclude
, prefix
, x
->patternlen
,
721 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
722 if (match_pathname(pathname
, pathlen
,
723 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
724 exclude
, prefix
, x
->patternlen
, x
->flags
))
727 return NULL
; /* undecided */
731 * Scan the list and let the last match determine the fate.
732 * Return 1 for exclude, 0 for include and -1 for undecided.
734 int is_excluded_from_list(const char *pathname
,
735 int pathlen
, const char *basename
, int *dtype
,
736 struct exclude_list
*el
)
738 struct exclude
*exclude
;
739 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
, dtype
, el
);
741 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
742 return -1; /* undecided */
745 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
746 const char *pathname
, int pathlen
, const char *basename
,
750 struct exclude_list_group
*group
;
751 struct exclude
*exclude
;
752 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
753 group
= &dir
->exclude_list_group
[i
];
754 for (j
= group
->nr
- 1; j
>= 0; j
--) {
755 exclude
= last_exclude_matching_from_list(
756 pathname
, pathlen
, basename
, dtype_p
,
766 * Loads the per-directory exclude list for the substring of base
767 * which has a char length of baselen.
769 static void prep_exclude(struct dir_struct
*dir
, const char *base
, int baselen
)
771 struct exclude_list_group
*group
;
772 struct exclude_list
*el
;
773 struct exclude_stack
*stk
= NULL
;
776 group
= &dir
->exclude_list_group
[EXC_DIRS
];
778 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
779 * which originate from directories not in the prefix of the
780 * path being checked. */
781 while ((stk
= dir
->exclude_stack
) != NULL
) {
782 if (stk
->baselen
<= baselen
&&
783 !strncmp(dir
->basebuf
, base
, stk
->baselen
))
785 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
786 dir
->exclude_stack
= stk
->prev
;
788 free((char *)el
->src
); /* see strdup() below */
789 clear_exclude_list(el
);
794 /* Skip traversing into sub directories if the parent is excluded */
798 /* Read from the parent directories and push them down. */
799 current
= stk
? stk
->baselen
: -1;
800 while (current
< baselen
) {
801 struct exclude_stack
*stk
= xcalloc(1, sizeof(*stk
));
809 cp
= strchr(base
+ current
+ 1, '/');
811 die("oops in prep_exclude");
814 stk
->prev
= dir
->exclude_stack
;
815 stk
->baselen
= cp
- base
;
816 stk
->exclude_ix
= group
->nr
;
817 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
818 memcpy(dir
->basebuf
+ current
, base
+ current
,
819 stk
->baselen
- current
);
821 /* Abort if the directory is excluded */
824 dir
->basebuf
[stk
->baselen
- 1] = 0;
825 dir
->exclude
= last_exclude_matching_from_lists(dir
,
826 dir
->basebuf
, stk
->baselen
- 1,
827 dir
->basebuf
+ current
, &dt
);
828 dir
->basebuf
[stk
->baselen
- 1] = '/';
830 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
833 dir
->basebuf
[stk
->baselen
] = 0;
834 dir
->exclude_stack
= stk
;
839 /* Try to read per-directory file unless path is too long */
840 if (dir
->exclude_per_dir
&&
841 stk
->baselen
+ strlen(dir
->exclude_per_dir
) < PATH_MAX
) {
842 strcpy(dir
->basebuf
+ stk
->baselen
,
843 dir
->exclude_per_dir
);
845 * dir->basebuf gets reused by the traversal, but we
846 * need fname to remain unchanged to ensure the src
847 * member of each struct exclude correctly
848 * back-references its source file. Other invocations
849 * of add_exclude_list provide stable strings, so we
850 * strdup() and free() here in the caller.
852 el
->src
= strdup(dir
->basebuf
);
853 add_excludes_from_file_to_list(dir
->basebuf
,
854 dir
->basebuf
, stk
->baselen
, el
, 1);
856 dir
->exclude_stack
= stk
;
857 current
= stk
->baselen
;
859 dir
->basebuf
[baselen
] = '\0';
863 * Loads the exclude lists for the directory containing pathname, then
864 * scans all exclude lists to determine whether pathname is excluded.
865 * Returns the exclude_list element which matched, or NULL for
868 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
869 const char *pathname
,
872 int pathlen
= strlen(pathname
);
873 const char *basename
= strrchr(pathname
, '/');
874 basename
= (basename
) ? basename
+1 : pathname
;
876 prep_exclude(dir
, pathname
, basename
-pathname
);
881 return last_exclude_matching_from_lists(dir
, pathname
, pathlen
,
886 * Loads the exclude lists for the directory containing pathname, then
887 * scans all exclude lists to determine whether pathname is excluded.
888 * Returns 1 if true, otherwise 0.
890 int is_excluded(struct dir_struct
*dir
, const char *pathname
, int *dtype_p
)
892 struct exclude
*exclude
=
893 last_exclude_matching(dir
, pathname
, dtype_p
);
895 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
899 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
901 struct dir_entry
*ent
;
903 ent
= xmalloc(sizeof(*ent
) + len
+ 1);
905 memcpy(ent
->name
, pathname
, len
);
910 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
, const char *pathname
, int len
)
912 if (cache_name_exists(pathname
, len
, ignore_case
))
915 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
916 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
919 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
, const char *pathname
, int len
)
921 if (!cache_name_is_other(pathname
, len
))
924 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
925 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
929 index_nonexistent
= 0,
935 * Do not use the alphabetically stored index to look up
936 * the directory name; instead, use the case insensitive
939 static enum exist_status
directory_exists_in_index_icase(const char *dirname
, int len
)
941 struct cache_entry
*ce
= index_name_exists(&the_index
, dirname
, len
+ 1, ignore_case
);
942 unsigned char endchar
;
945 return index_nonexistent
;
946 endchar
= ce
->name
[len
];
949 * The cache_entry structure returned will contain this dirname
950 * and possibly additional path components.
953 return index_directory
;
956 * If there are no additional path components, then this cache_entry
957 * represents a submodule. Submodules, despite being directories,
958 * are stored in the cache without a closing slash.
960 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
963 /* This should never be hit, but it exists just in case. */
964 return index_nonexistent
;
968 * The index sorts alphabetically by entry name, which
969 * means that a gitlink sorts as '\0' at the end, while
970 * a directory (which is defined not as an entry, but as
971 * the files it contains) will sort with the '/' at the
974 static enum exist_status
directory_exists_in_index(const char *dirname
, int len
)
979 return directory_exists_in_index_icase(dirname
, len
);
981 pos
= cache_name_pos(dirname
, len
);
984 while (pos
< active_nr
) {
985 struct cache_entry
*ce
= active_cache
[pos
++];
986 unsigned char endchar
;
988 if (strncmp(ce
->name
, dirname
, len
))
990 endchar
= ce
->name
[len
];
994 return index_directory
;
995 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
998 return index_nonexistent
;
1002 * When we find a directory when traversing the filesystem, we
1003 * have three distinct cases:
1006 * - see it as a directory
1009 * and which one we choose depends on a combination of existing
1010 * git index contents and the flags passed into the directory
1011 * traversal routine.
1013 * Case 1: If we *already* have entries in the index under that
1014 * directory name, we always recurse into the directory to see
1017 * Case 2: If we *already* have that directory name as a gitlink,
1018 * we always continue to see it as a gitlink, regardless of whether
1019 * there is an actual git directory there or not (it might not
1020 * be checked out as a subproject!)
1022 * Case 3: if we didn't have it in the index previously, we
1023 * have a few sub-cases:
1025 * (a) if "show_other_directories" is true, we show it as
1026 * just a directory, unless "hide_empty_directories" is
1027 * also true, in which case we need to check if it contains any
1028 * untracked and / or ignored files.
1029 * (b) if it looks like a git directory, and we don't have
1030 * 'no_gitlinks' set we treat it as a gitlink, and show it
1032 * (c) otherwise, we recurse into it.
1034 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1035 const char *dirname
, int len
, int exclude
,
1036 const struct path_simplify
*simplify
)
1038 /* The "len-1" is to strip the final '/' */
1039 switch (directory_exists_in_index(dirname
, len
-1)) {
1040 case index_directory
:
1041 return path_recurse
;
1046 case index_nonexistent
:
1047 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1049 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1050 unsigned char sha1
[20];
1051 if (resolve_gitlink_ref(dirname
, "HEAD", sha1
) == 0)
1052 return path_untracked
;
1054 return path_recurse
;
1057 /* This is the "show_other_directories" case */
1059 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1060 return exclude
? path_excluded
: path_untracked
;
1062 return read_directory_recursive(dir
, dirname
, len
, 1, simplify
);
1066 * This is an inexact early pruning of any recursive directory
1067 * reading - if the path cannot possibly be in the pathspec,
1068 * return true, and we'll skip it early.
1070 static int simplify_away(const char *path
, int pathlen
, const struct path_simplify
*simplify
)
1074 const char *match
= simplify
->path
;
1075 int len
= simplify
->len
;
1081 if (!memcmp(path
, match
, len
))
1091 * This function tells us whether an excluded path matches a
1092 * list of "interesting" pathspecs. That is, whether a path matched
1093 * by any of the pathspecs could possibly be ignored by excluding
1094 * the specified path. This can happen if:
1096 * 1. the path is mentioned explicitly in the pathspec
1098 * 2. the path is a directory prefix of some element in the
1101 static int exclude_matches_pathspec(const char *path
, int len
,
1102 const struct path_simplify
*simplify
)
1105 for (; simplify
->path
; simplify
++) {
1106 if (len
== simplify
->len
1107 && !memcmp(path
, simplify
->path
, len
))
1109 if (len
< simplify
->len
1110 && simplify
->path
[len
] == '/'
1111 && !memcmp(path
, simplify
->path
, len
))
1118 static int get_index_dtype(const char *path
, int len
)
1121 struct cache_entry
*ce
;
1123 ce
= cache_name_exists(path
, len
, 0);
1125 if (!ce_uptodate(ce
))
1127 if (S_ISGITLINK(ce
->ce_mode
))
1130 * Nobody actually cares about the
1131 * difference between DT_LNK and DT_REG
1136 /* Try to look it up as a directory */
1137 pos
= cache_name_pos(path
, len
);
1141 while (pos
< active_nr
) {
1142 ce
= active_cache
[pos
++];
1143 if (strncmp(ce
->name
, path
, len
))
1145 if (ce
->name
[len
] > '/')
1147 if (ce
->name
[len
] < '/')
1149 if (!ce_uptodate(ce
))
1150 break; /* continue? */
1156 static int get_dtype(struct dirent
*de
, const char *path
, int len
)
1158 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1161 if (dtype
!= DT_UNKNOWN
)
1163 dtype
= get_index_dtype(path
, len
);
1164 if (dtype
!= DT_UNKNOWN
)
1166 if (lstat(path
, &st
))
1168 if (S_ISREG(st
.st_mode
))
1170 if (S_ISDIR(st
.st_mode
))
1172 if (S_ISLNK(st
.st_mode
))
1177 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1178 struct strbuf
*path
,
1179 const struct path_simplify
*simplify
,
1180 int dtype
, struct dirent
*de
)
1183 if (dtype
== DT_UNKNOWN
)
1184 dtype
= get_dtype(de
, path
->buf
, path
->len
);
1186 /* Always exclude indexed files */
1187 if (dtype
!= DT_DIR
&&
1188 cache_name_exists(path
->buf
, path
->len
, ignore_case
))
1191 exclude
= is_excluded(dir
, path
->buf
, &dtype
);
1194 * Excluded? If we don't explicitly want to show
1195 * ignored files, ignore it
1197 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1198 return path_excluded
;
1204 strbuf_addch(path
, '/');
1205 return treat_directory(dir
, path
->buf
, path
->len
, exclude
,
1209 return exclude
? path_excluded
: path_untracked
;
1213 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1215 struct strbuf
*path
,
1217 const struct path_simplify
*simplify
)
1221 if (is_dot_or_dotdot(de
->d_name
) || !strcmp(de
->d_name
, ".git"))
1223 strbuf_setlen(path
, baselen
);
1224 strbuf_addstr(path
, de
->d_name
);
1225 if (simplify_away(path
->buf
, path
->len
, simplify
))
1229 return treat_one_path(dir
, path
, simplify
, dtype
, de
);
1233 * Read a directory tree. We currently ignore anything but
1234 * directories, regular files and symlinks. That's because git
1235 * doesn't handle them at all yet. Maybe that will change some
1238 * Also, we ignore the name ".git" (even if it is not a directory).
1239 * That likely will not change.
1241 * Returns the most significant path_treatment value encountered in the scan.
1243 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1244 const char *base
, int baselen
,
1246 const struct path_simplify
*simplify
)
1249 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1251 struct strbuf path
= STRBUF_INIT
;
1253 strbuf_add(&path
, base
, baselen
);
1255 fdir
= opendir(path
.len
? path
.buf
: ".");
1259 while ((de
= readdir(fdir
)) != NULL
) {
1260 /* check how the file or directory should be treated */
1261 state
= treat_path(dir
, de
, &path
, baselen
, simplify
);
1262 if (state
> dir_state
)
1265 /* recurse into subdir if instructed by treat_path */
1266 if (state
== path_recurse
) {
1267 subdir_state
= read_directory_recursive(dir
, path
.buf
,
1268 path
.len
, check_only
, simplify
);
1269 if (subdir_state
> dir_state
)
1270 dir_state
= subdir_state
;
1274 /* abort early if maximum state has been reached */
1275 if (dir_state
== path_untracked
)
1277 /* skip the dir_add_* part */
1281 /* add the path to the appropriate result list */
1284 if (dir
->flags
& DIR_SHOW_IGNORED
)
1285 dir_add_name(dir
, path
.buf
, path
.len
);
1286 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1287 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1288 exclude_matches_pathspec(path
.buf
, path
.len
,
1290 dir_add_ignored(dir
, path
.buf
, path
.len
);
1293 case path_untracked
:
1294 if (!(dir
->flags
& DIR_SHOW_IGNORED
))
1295 dir_add_name(dir
, path
.buf
, path
.len
);
1304 strbuf_release(&path
);
1309 static int cmp_name(const void *p1
, const void *p2
)
1311 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
1312 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
1314 return cache_name_compare(e1
->name
, e1
->len
,
1318 static struct path_simplify
*create_simplify(const char **pathspec
)
1321 struct path_simplify
*simplify
= NULL
;
1326 for (nr
= 0 ; ; nr
++) {
1329 alloc
= alloc_nr(alloc
);
1330 simplify
= xrealloc(simplify
, alloc
* sizeof(*simplify
));
1332 match
= *pathspec
++;
1335 simplify
[nr
].path
= match
;
1336 simplify
[nr
].len
= simple_length(match
);
1338 simplify
[nr
].path
= NULL
;
1339 simplify
[nr
].len
= 0;
1343 static void free_simplify(struct path_simplify
*simplify
)
1348 static int treat_leading_path(struct dir_struct
*dir
,
1349 const char *path
, int len
,
1350 const struct path_simplify
*simplify
)
1352 struct strbuf sb
= STRBUF_INIT
;
1353 int baselen
, rc
= 0;
1355 int old_flags
= dir
->flags
;
1357 while (len
&& path
[len
- 1] == '/')
1362 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
1364 cp
= path
+ baselen
+ !!baselen
;
1365 cp
= memchr(cp
, '/', path
+ len
- cp
);
1369 baselen
= cp
- path
;
1370 strbuf_setlen(&sb
, 0);
1371 strbuf_add(&sb
, path
, baselen
);
1372 if (!is_directory(sb
.buf
))
1374 if (simplify_away(sb
.buf
, sb
.len
, simplify
))
1376 if (treat_one_path(dir
, &sb
, simplify
,
1377 DT_DIR
, NULL
) == path_none
)
1378 break; /* do not recurse into it */
1379 if (len
<= baselen
) {
1381 break; /* finished checking */
1384 strbuf_release(&sb
);
1385 dir
->flags
= old_flags
;
1389 int read_directory(struct dir_struct
*dir
, const char *path
, int len
, const char **pathspec
)
1391 struct path_simplify
*simplify
;
1393 if (has_symlink_leading_path(path
, len
))
1396 simplify
= create_simplify(pathspec
);
1397 if (!len
|| treat_leading_path(dir
, path
, len
, simplify
))
1398 read_directory_recursive(dir
, path
, len
, 0, simplify
);
1399 free_simplify(simplify
);
1400 qsort(dir
->entries
, dir
->nr
, sizeof(struct dir_entry
*), cmp_name
);
1401 qsort(dir
->ignored
, dir
->ignored_nr
, sizeof(struct dir_entry
*), cmp_name
);
1405 int file_exists(const char *f
)
1408 return lstat(f
, &sb
) == 0;
1412 * Given two normalized paths (a trailing slash is ok), if subdir is
1413 * outside dir, return -1. Otherwise return the offset in subdir that
1414 * can be used as relative path to dir.
1416 int dir_inside_of(const char *subdir
, const char *dir
)
1420 assert(dir
&& subdir
&& *dir
&& *subdir
);
1422 while (*dir
&& *subdir
&& *dir
== *subdir
) {
1428 /* hel[p]/me vs hel[l]/yeah */
1429 if (*dir
&& *subdir
)
1433 return !*dir
? offset
: -1; /* same dir */
1435 /* foo/[b]ar vs foo/[] */
1436 if (is_dir_sep(dir
[-1]))
1437 return is_dir_sep(subdir
[-1]) ? offset
: -1;
1439 /* foo[/]bar vs foo[] */
1440 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
1443 int is_inside_dir(const char *dir
)
1448 if (!getcwd(cwd
, sizeof(cwd
)))
1449 die_errno("can't find the current directory");
1450 return dir_inside_of(cwd
, dir
) >= 0;
1453 int is_empty_dir(const char *path
)
1455 DIR *dir
= opendir(path
);
1462 while ((e
= readdir(dir
)) != NULL
)
1463 if (!is_dot_or_dotdot(e
->d_name
)) {
1472 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
1476 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
1477 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
1478 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
1479 unsigned char submodule_head
[20];
1481 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
1482 !resolve_gitlink_ref(path
->buf
, "HEAD", submodule_head
)) {
1483 /* Do not descend and nuke a nested git work tree. */
1489 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
1490 dir
= opendir(path
->buf
);
1492 /* an empty dir could be removed even if it is unreadble */
1494 return rmdir(path
->buf
);
1498 if (path
->buf
[original_len
- 1] != '/')
1499 strbuf_addch(path
, '/');
1502 while ((e
= readdir(dir
)) != NULL
) {
1504 if (is_dot_or_dotdot(e
->d_name
))
1507 strbuf_setlen(path
, len
);
1508 strbuf_addstr(path
, e
->d_name
);
1509 if (lstat(path
->buf
, &st
))
1511 else if (S_ISDIR(st
.st_mode
)) {
1512 if (!remove_dir_recurse(path
, flag
, &kept_down
))
1513 continue; /* happy */
1514 } else if (!only_empty
&& !unlink(path
->buf
))
1515 continue; /* happy, too */
1517 /* path too long, stat fails, or non-directory still exists */
1523 strbuf_setlen(path
, original_len
);
1524 if (!ret
&& !keep_toplevel
&& !kept_down
)
1525 ret
= rmdir(path
->buf
);
1528 * report the uplevel that it is not an error that we
1529 * did not rmdir() our directory.
1535 int remove_dir_recursively(struct strbuf
*path
, int flag
)
1537 return remove_dir_recurse(path
, flag
, NULL
);
1540 void setup_standard_excludes(struct dir_struct
*dir
)
1545 dir
->exclude_per_dir
= ".gitignore";
1546 path
= git_path("info/exclude");
1547 if (!excludes_file
) {
1548 home_config_paths(NULL
, &xdg_path
, "ignore");
1549 excludes_file
= xdg_path
;
1551 if (!access_or_warn(path
, R_OK
, 0))
1552 add_excludes_from_file(dir
, path
);
1553 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
1554 add_excludes_from_file(dir
, excludes_file
);
1557 int remove_path(const char *name
)
1561 if (unlink(name
) && errno
!= ENOENT
&& errno
!= ENOTDIR
)
1564 slash
= strrchr(name
, '/');
1566 char *dirs
= xstrdup(name
);
1567 slash
= dirs
+ (slash
- name
);
1570 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
1576 static int pathspec_item_cmp(const void *a_
, const void *b_
)
1578 struct pathspec_item
*a
, *b
;
1580 a
= (struct pathspec_item
*)a_
;
1581 b
= (struct pathspec_item
*)b_
;
1582 return strcmp(a
->match
, b
->match
);
1585 int init_pathspec(struct pathspec
*pathspec
, const char **paths
)
1587 const char **p
= paths
;
1590 memset(pathspec
, 0, sizeof(*pathspec
));
1595 pathspec
->raw
= paths
;
1596 pathspec
->nr
= p
- paths
;
1600 pathspec
->items
= xmalloc(sizeof(struct pathspec_item
)*pathspec
->nr
);
1601 for (i
= 0; i
< pathspec
->nr
; i
++) {
1602 struct pathspec_item
*item
= pathspec
->items
+i
;
1603 const char *path
= paths
[i
];
1606 item
->original
= path
;
1607 item
->len
= strlen(path
);
1609 if (limit_pathspec_to_literal()) {
1610 item
->nowildcard_len
= item
->len
;
1612 item
->nowildcard_len
= simple_length(path
);
1613 if (item
->nowildcard_len
< item
->len
) {
1614 pathspec
->has_wildcard
= 1;
1615 if (path
[item
->nowildcard_len
] == '*' &&
1616 no_wildcard(path
+ item
->nowildcard_len
+ 1))
1617 item
->flags
|= PATHSPEC_ONESTAR
;
1622 qsort(pathspec
->items
, pathspec
->nr
,
1623 sizeof(struct pathspec_item
), pathspec_item_cmp
);
1628 void free_pathspec(struct pathspec
*pathspec
)
1630 free(pathspec
->items
);
1631 pathspec
->items
= NULL
;
1634 int limit_pathspec_to_literal(void)
1636 static int flag
= -1;
1638 flag
= git_env_bool(GIT_LITERAL_PATHSPECS_ENVIRONMENT
, 0);
1643 * Frees memory within dir which was allocated for exclude lists and
1644 * the exclude_stack. Does not free dir itself.
1646 void clear_directory(struct dir_struct
*dir
)
1649 struct exclude_list_group
*group
;
1650 struct exclude_list
*el
;
1651 struct exclude_stack
*stk
;
1653 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1654 group
= &dir
->exclude_list_group
[i
];
1655 for (j
= 0; j
< group
->nr
; j
++) {
1658 free((char *)el
->src
);
1659 clear_exclude_list(el
);
1664 stk
= dir
->exclude_stack
;
1666 struct exclude_stack
*prev
= stk
->prev
;